Visual motion testing system based on artificial intelligence algorithm

Abstract

The present invention discloses a visual motion test system based on an artificial intelligence algorithm, comprising a motion video acquisition module, a motion action analysis module, an abnormality verification and analysis module, a motion test scoring module and a test information output module. The present invention relates to the field of motion test technology, and solves the technical problem that the motion video acquired by a machine may cause differences due to the influence of external factors, thereby affecting subsequent recognition and judgment, and causing errors in the scoring of athletes. The present invention segments the motion video acquired by the machine, and performs image segmentation processing on the segmented motion video, and determines whether there is an abnormal action according to the action of the athlete in the segmented image, and further analyzes the influence of the abnormal action, specifically from the two aspects of image shadow and image light, and adjusts the video as a whole according to the specific reasons obtained, thereby reducing the error of subsequent analysis and improving the accuracy of the analysis.

Description

Technical Field

[0001] The present invention relates to the technical field of motion testing, and in particular to a visual motion testing system based on an artificial intelligence algorithm. Background Art

[0002] In modern society, sports are essential for the healthy growth and all-round development of primary and secondary school students. In order to provide better physical education and exercise, schools will conduct corresponding physical education examinations and record the sports performance of students in their majors based on the examinations.

[0003] According to the application number CN202311344347.3, a method and system for multi-person motion testing based on visual technology are disclosed, and the method includes: fixing the camera at a set height, and then setting multiple test points and a main control position according to the current video stream of the camera; the control personnel at the main control position selects a sports event, and the corresponding picture is displayed on the screen; each tester enters the corresponding test point, and after all testers are ready, the screen displays the entry motion detection picture; during the detection, the corresponding motion is detected by obtaining the coordinates of the skeleton points, and the corresponding real-time test information is provided for each tester on the screen; after the test is completed, the corresponding test results of each tester are displayed; it is convenient to conduct multi-person testing and improve the test efficiency.

[0004] When some existing visual motion testing systems are used, they obtain the corresponding motion video through a machine and analyze the motion video for testing. However, when the machine obtains the motion video, it may be affected by external factors, and the obtained video may have different differences, which may further lead to recognition errors in the subsequent recognition of the athlete's movements, thereby affecting the athlete's score. Summary of the invention

[0005] In view of the deficiencies in the prior art, the present invention provides a visual motion testing system based on an artificial intelligence algorithm, which solves the problem that the sports videos acquired by the machine may be different due to the influence of external factors, thereby affecting subsequent recognition and judgment, and causing errors in the athletes' scores.

[0006] To achieve the above objectives, the present invention is implemented through the following technical solutions: a visual motion test system based on artificial intelligence algorithm, comprising:

[0007] A motion video acquisition module, used for transmitting the acquired motion video to the motion action analysis module;

[0008] The motion analysis module is used to analyze the acquired motion video, and to segment the motion video according to time to obtain segmented videos, and to compare the motions in the segmented videos for abnormalities and obtain comparison results, wherein the comparison results include normal motion signals and abnormal motion signals, and to transmit the abnormal motion signals in the comparison results to the abnormality verification analysis module;

[0009] The abnormal verification analysis module is used to analyze the acquired abnormal motion signal, generate the impact result and the non-impact result by performing the impact analysis on the segmented image background corresponding to the abnormal motion signal, and perform the secondary analysis on the impact result to obtain the specific impact factor, and perform the secondary identification according to the specific impact factor to obtain the verification information, wherein the verification information includes the normal verification result and the abnormal verification result, and transmit the normal verification result to the motion test scoring module;

[0010] The sports test scoring module is used to analyze the normal verification results obtained, generate correction information by correcting the abnormal movements obtained, and score the athletes in combination with standard movements, and obtain the scoring results at the same time, and then transmit the scoring results to the test information output module.

[0011] The test information output module is used to display the obtained correction information and scoring results to the corresponding operators.

[0012] As a further solution of the present invention: the specific method of generating the comparison result by the motion analysis module is:

[0013] A motion video is obtained, and the motion video is segmented with time t as the segmentation point to obtain segmented videos, and the segmented videos are labeled as i, and i=1, 2, ..., j, where j represents the number of segmented videos, and then the segmented video corresponding to the label i=1 is obtained as the target object, and the target object is segmented to obtain a segmented image, and the action of the moving person in the segmented image is recognized;

[0014] The motion action corresponding to the moving person in the segmented image is obtained, and the corresponding motion trajectory is generated according to the motion action. Then the motion trajectory is compared and matched with the standard action. When the motion trajectory matches the standard action, a normal motion signal is generated. When the motion trajectory does not match the standard action, an abnormal motion signal is generated.

[0015] As a further solution of the present invention: the specific manner in which the abnormality verification analysis module performs an impact analysis on the segmented image background to generate impact results and non-impact results is:

[0016] The segmented image corresponding to the abnormal action signal is obtained and recorded as the image to be analyzed, and the segmented image corresponding to the normal action is obtained and recorded as the normal image. At the same time, the normal image is compared with the image to be analyzed. If the normal image has the same background as the image to be analyzed, it means that the background in the image to be analyzed has no influence, and a result without influence is generated. On the contrary, if the normal image does not have the same background as the image to be analyzed, it means that the background in the image to be analyzed has an influence, and an influence result is generated, and a secondary analysis is performed on the influence result.

[0017] As a further solution of the present invention, the specific method of the abnormal verification and analysis module to perform secondary analysis on the impact results is:

[0018] The shadows and lights in the image to be analyzed are analyzed separately. When analyzing the shadows, the light conditions are controlled to remain unchanged for analysis. If changing the shadows has no effect on the recognition of the image to be analyzed, a normal verification result is generated. Conversely, if changing the shadows affects the recognition of the image to be analyzed, an abnormal verification result is generated. Similarly, the light is analyzed by the controlled variable method to obtain verification information, and the segmented image is adjusted according to the verification information.

[0019] As a further solution of the present invention, the specific method in which the abnormal verification analysis module adjusts the segmented image according to the verification information is as follows:

[0020] The segmented image is adjusted based on the verification information to obtain the segmented image corresponding to the normal action and the corresponding influencing parameter information in the segmented image, and the influencing parameter information includes: shadow parameters and light parameters. At the same time, the segmented image corresponding to the abnormal action is adjusted according to the influencing parameters to obtain a processed image, and the processed image is transmitted to the motion test scoring module.

[0021] As a further solution of the present invention: the specific method of generating the correction information by the motion test scoring module is:

[0022] The abnormal action corresponding to the segmented image is obtained, and the abnormal action is compared with the standard action, the difference between the two is calculated, and correction information is generated according to the difference.

[0023] As a further solution of the present invention: the specific method of generating the scoring result by the sports test scoring module is:

[0024] Get the scoring interval, match the difference with the scoring interval and get the scoring value corresponding to the abnormal action, then calculate the average score corresponding to all segmented images, and use the average score as the standard value of the segmented video;

[0025] In this way, standard values ​​are obtained for all segmented videos, and the mean of the standard values ​​is calculated. The calculated mean of the standard values ​​is used as the scoring standard for the athletes, and the scoring results are generated, and then transmitted to the test information output module.

[0026] The present invention provides a visual motion test system based on artificial intelligence algorithm. Compared with the prior art, it has the following beneficial effects:

[0027] The present invention segments the sports video acquired by a machine, and performs image segmentation processing on the segmented sports video. It determines whether there are abnormal actions based on the actions of athletes in the segmented image, and further analyzes the impact of the abnormal actions, specifically from two aspects: image shadow and image light. The video as a whole is adjusted according to the specific reasons obtained, thereby reducing the error of subsequent analysis and improving the accuracy of analysis. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] Figure 1 This is a block diagram of the system principle of the present invention. DETAILED DESCRIPTION

[0029] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0030] For example, see Figure 1 The present application provides a visual motion testing system based on an artificial intelligence algorithm, including a motion video acquisition module, a motion action analysis module, an abnormality verification analysis module, a motion test scoring module and a test information output module, and each of the above units is electrically connected in a unidirectional manner.

[0031] The sports video acquisition module is used to acquire the sports videos of athletes and transmit the acquired sports videos to the sports action analysis module. The acquisition method here is to acquire the sports videos of athletes through a high-precision camera.

[0032] The motion analysis module is used to analyze the acquired motion video, and to segment the motion video according to time to obtain segmented videos. At the same time, the motion actions in the segmented videos are compared for abnormality and a comparison result is obtained, wherein the comparison result includes a normal motion signal and an abnormal motion signal. At the same time, the abnormal motion signal in the comparison result is transmitted to the abnormal verification analysis module, and the specific method of generating the comparison result is:

[0033] Obtain a motion video, and segment the motion video with time t as the segmentation point to obtain a segmented video, and the specific value of time t here is set by the operator, and the segmented video is labeled as i, and i=1, 2, ..., j, where j represents the number of segmented videos, then obtain the segmented video corresponding to the label i=1 as the target object, and perform image segmentation on the target object to obtain a segmented image, and the segmented image here represents a unit time image, specifically expressed as a precise image at a certain moment, and at the same time identify the action of the sports person in the segmented image; these segmented images represent the precise picture within a unit time, and each frame corresponds to the exact visual information at a certain moment. Further action recognition analysis of these segmented images can accurately identify the specific actions of the sports person.

[0034] The motion action corresponding to the moving person in the segmented image is obtained, and the corresponding motion trajectory is generated according to the motion action. The motion trajectory generated here is to obtain the standard action at the same time, and then compare and match the motion trajectory with the standard action. When the motion trajectory matches the standard action, a normal action signal is generated. When the motion trajectory does not match the standard action, an abnormal action signal is generated. Specifically, the match here means that the difference between the motion trajectory and the standard action is within the allowable range. At the same time, the allowable range of the difference is set by the operator. This allowable range is set by the operator according to actual needs, and it defines what is a "standard" or "normal" action. On the contrary, if the motion trajectory deviates from the standard action beyond the allowable range, the system will generate an "abnormal action signal".

[0035] Consider a gymnast who is training for the balance beam event. The coach wants to ensure that the athlete's performance of a specific move, such as a backflip, is up to international competition standards. First, the instantaneous backflip movement is extracted from the training video using the video segmentation and image processing steps described previously, and the motion trajectory of this move is generated;

[0036] The system has stored the standard movement trajectory of backflips, which is obtained by averaging the movements of winning athletes in multiple world-class competitions. The training system compares the athlete's movement trajectory with this standard trajectory. If the deviation between the athlete's movement trajectory and the standard trajectory is within the allowable range set by the coach (for example, flip angle, height, completion time, etc.), the system determines that the backflip is a "normal action", and the corresponding signal will inform the coach that the action is in line with expectations. If the deviation between the athlete's backflip movement trajectory and the standard trajectory exceeds the allowable range, such as the flip axis offset is too large or the steps are unstable when landing, the system will generate an "abnormal action signal".

[0037] The abnormal verification analysis module is used to analyze the acquired abnormal motion signal, generate the impact result and the non-impact result by performing the impact analysis on the segmented image background corresponding to the abnormal motion signal, and perform a secondary analysis on the impact result to obtain the specific impact factor, and perform a secondary identification based on the specific impact factor to obtain the verification information, wherein the verification information includes the normal verification result and the abnormal verification result, and transmits the normal verification result to the motion test scoring module, and the specific method of generating the verification information is:

[0038] A segmented image corresponding to the abnormal action signal is obtained and recorded as the image to be analyzed, and at the same time, an impact judgment is performed on the background in the image to be analyzed, and a specific impact judgment method is as follows: a segmented image corresponding to a normal action is obtained and recorded as a normal image, and at the same time, the normal image is compared with the image to be analyzed. If the normal image has the same background as the image to be analyzed, it means that the background in the image to be analyzed has no impact, and a no-impact result is generated. Conversely, if the normal image does not have the same background as the image to be analyzed, it means that the background in the image to be analyzed has an impact, and an impact result is generated.

[0039] Then, a secondary analysis is performed on the generated impact results, and the shadows and lights in the image to be analyzed are analyzed separately, and the control variable method is used for analysis. Specifically, when analyzing the shadow, the light in the image to be analyzed is controlled to remain unchanged, and the impact of the shadow on the image to be analyzed is judged by changing the shadow situation for analysis. If the recognition of the image to be analyzed has no effect in the process of changing the shadow, a normal verification result is generated. On the contrary, if the process of changing the shadow affects the recognition of the image to be analyzed, an abnormal verification result is generated. Specifically, if the shadow change causes the initial abnormal action recognition to become a normal action, it means that the shadow has an impact on the recognition of the image to be analyzed. If the shadow change is always recognized as an abnormal action, it means that the shadow has no impact on the recognition of the image to be analyzed. Similarly, the light is analyzed by the control variable method to obtain the corresponding verification information.

[0040] Consider a basketball training scenario where a player is practicing free throws. Assume that on a cloudy day, the system sends out an abnormal action signal, indicating that the player's free throw action does not meet the standard. For further analysis, the coach decides to use the control variable method mentioned above.

[0041] First, to analyze the impact of shadows, the coach records the video at different time points to observe the impact of different shadow positions under the same lighting conditions. If it is found that when the shadow covers the basketball hoop, the system mistakenly identifies a normal free throw as an anomaly, then it can be concluded that the presence of shadows has an impact on the recognition of the image to be analyzed.

[0042] The instructor then adjusts the lighting conditions, such as using additional lights to simulate different lighting environments. If the system's recognition results change from abnormal to normal after increasing the light intensity, this indicates that the lighting conditions are a key factor affecting accurate recognition.

[0043] The segmented image is adjusted based on the verification information to obtain the segmented image corresponding to the normal action, and the corresponding influencing parameter information in the segmented image is obtained, and the influencing parameter information includes: shadow parameters and light parameters. At the same time, the segmented image corresponding to the abnormal action is adjusted according to the influencing parameters to obtain a processed image, and the processed image is transmitted to the motion test scoring module at the same time; and the segmented image corresponding to the abnormal action here is represented as a segmented image in which the action recognition is abnormal due to shadow or light reasons, and the abnormal action not caused by the two is not obtained.

[0044] For abnormal verification results in the verification information, the system processes the segmented image according to the corresponding specific influencing factors. For example, if the shadow in the segmented image causes abnormalities in the overall motion recognition, the shadow parameters corresponding to the normal segmented image are selected and adjusted as the standard. The light analysis method is similar. At the same time, the processed segmented image is transmitted to the motion test scoring module. Secondly, when shooting later, the captured image is adjusted according to the specific influencing factors.

[0045] The sports test scoring module is used to analyze the obtained normal verification results, generate correction information by correcting the abnormal movements, and score the athletes in combination with standard movements, and obtain the scoring results at the same time, and then transmit the scoring results to the test information output module. The specific method of obtaining the scoring results is:

[0046] The abnormal action corresponding to the segmented image is obtained, and the abnormal action is compared with the standard action, and the difference between the two is calculated. At the same time, correction information is generated according to the difference. The correction information generated here is obtained by calculating and analyzing the corresponding parameters by a computer, and using an artificial intelligence algorithm to analyze the correction information, and the correction information is transmitted to the test information output module. Then, the scoring interval is obtained, and the scoring interval here is represented by the scoring value corresponding to different actions, and the specific scoring standard of the scoring interval is set by the operator. At the same time, the difference is matched with the scoring interval and the scoring value corresponding to the abnormal action is obtained. The scoring value obtained here is the scoring value corresponding to the segmented image. Then, the average score corresponding to all segmented images is calculated, and the average score is used as the standard value of the segmented video. Similarly, the standard value of all segmented videos is obtained, and the average of the standard value is calculated. The calculated average standard value is used as the scoring standard for the athlete, and the scoring result is generated, and then the scoring result is transmitted to the test information output module.

[0047] Suppose a gymnast is training on the uneven bars, and the coach wants to accurately score her movements. First, the system captures and segments a series of images of the athlete's movements and compares them with the movements in the standard movement library. If a certain movement, such as insufficient rotation, is found to be different from the standard movement, the system will calculate the specific value of the difference and generate correction information.

[0048] According to the pre-set scoring criteria, such as the degree of rotation corresponding to different score ranges, the system matches the difference with the scoring range and gives the score value of the action. After that, this process is repeated for all segmented images to obtain the score of each action and calculate the average.

[0049] Next, the average score of all segmented videos is averaged again to get the comprehensive score of the athlete's training. This score reflects the athlete's performance throughout the training process. The coach can adjust the training plan based on this result, and the athlete can also identify the details of the action that need to be improved.

[0050] A test information output module is used to display the obtained bisection results and correction information to the corresponding operators.

[0051] Some of the data in the above formulas are dimensionless and numerically calculated. Meanwhile, the contents not described in detail in this specification belong to the prior art known to those skilled in the art.

[0052] The above embodiments are only used to illustrate the technical method of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical method of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical method of the present invention.

Claims

1. A visual motion test system based on artificial intelligence algorithm, characterized in that: include: A motion video acquisition module, used for transmitting the acquired motion video to the motion action analysis module; The motion analysis module is used to analyze the acquired motion video, and to segment the motion video according to time to obtain segmented videos, and to compare the motions in the segmented videos for abnormalities and obtain comparison results, wherein the comparison results include normal motion signals and abnormal motion signals, and to transmit the abnormal motion signals in the comparison results to the abnormality verification analysis module; The abnormal verification analysis module is used to analyze the acquired abnormal motion signal, generate the impact result and the non-impact result by performing the impact analysis on the segmented image background corresponding to the abnormal motion signal, and perform a secondary analysis on the impact result to obtain the specific impact factor, and perform a secondary identification based on the specific impact factor to obtain the verification information, wherein the verification information includes the normal verification result and the abnormal verification result, and transmits the normal verification result to the motion test scoring module, and the specific processing method is as follows: The segmented image corresponding to the abnormal action signal is obtained and recorded as the image to be analyzed, and the segmented image corresponding to the normal action is obtained and recorded as the normal image. At the same time, the normal image is compared with the image to be analyzed. If the normal image has the same background as the image to be analyzed, it means that the background in the image to be analyzed has no influence, and a no-influence result is generated. On the contrary, if the normal image does not have the same background as the image to be analyzed, it means that the background in the image to be analyzed has an influence, and an influence result is generated, and a secondary analysis is performed on the influence result. The shadows and lights in the image to be analyzed are analyzed separately. When analyzing the shadows, the light conditions are controlled to remain unchanged. If changing the shadows has no effect on the recognition of the image to be analyzed, a normal verification result is generated. On the contrary, if changing the shadows affects the recognition of the image to be analyzed, an abnormal verification result is generated. Similarly, the light is analyzed by the controlled variable method to obtain verification information, and the segmented image is adjusted according to the verification information. The segmented image is adjusted based on the verification information, the segmented image corresponding to the normal action is obtained, and the corresponding influencing parameter information in the segmented image is obtained, and the influencing parameter information includes: shadow parameters and light parameters, and the segmented image corresponding to the abnormal action is adjusted according to the influencing parameters to obtain a processed image, and the processed image is transmitted to the motion test scoring module; The sports test scoring module is used to analyze the normal verification results obtained, generate correction information by correcting the abnormal movements obtained, and score the athletes in combination with standard movements, and obtain the scoring results at the same time, and then transmit the scoring results to the test information output module.

2. The visual motion testing system based on artificial intelligence algorithm according to claim 1 is characterized in that: It also includes a test information output module for displaying the obtained correction information and scoring results to the corresponding operators.

3. The visual motion testing system based on artificial intelligence algorithm according to claim 1 is characterized in that: The specific method of generating the comparison result by the motion analysis module is as follows: A motion video is obtained, and the motion video is segmented with time t as the segmentation point to obtain segmented videos, and the segmented videos are labeled as i, and i=1, 2, ..., j, where j represents the number of segmented videos, and then the segmented video corresponding to the label i=1 is obtained as the target object, and the target object is segmented to obtain a segmented image, and the action of the moving person in the segmented image is recognized; The motion action corresponding to the moving person in the segmented image is obtained, and the corresponding motion trajectory is generated according to the motion action. Then the motion trajectory is compared and matched with the standard action. When the motion trajectory matches the standard action, a normal motion signal is generated. When the motion trajectory does not match the standard action, an abnormal motion signal is generated.

4. The visual motion testing system based on artificial intelligence algorithm according to claim 1 is characterized in that: The specific method of generating the correction information by the motion test scoring module is as follows: The abnormal action corresponding to the segmented image is obtained, and the abnormal action is compared with the standard action, the difference between the two is calculated, and correction information is generated according to the difference.

5. The visual motion testing system based on artificial intelligence algorithm according to claim 1 is characterized in that: The specific method of generating the scoring result by the sports test scoring module is as follows: Get the scoring interval, match the difference with the scoring interval and get the scoring value corresponding to the abnormal action, then calculate the average score corresponding to all segmented images, and use the average score as the standard value of the segmented video; In this way, standard values ​​are obtained for all segmented videos, and the mean of the standard values ​​is calculated. The calculated mean of the standard values ​​is used as the scoring standard for the athletes, and the scoring results are generated, and then transmitted to the test information output module.

Images